AWS Summit New York 2026: Kiro Goes Mobile and DevOps Agent Gains Release Management
AWS held its Summit in New York on June 17, 2026, with the developer announcements centering on two themes: extending AI-native developer tooling to new form factors and automating the release lifecycle itself. The headline developer story is the Kiro for iOS gated preview, which brings a native mobile interface to AWS's AI-powered IDE assistant. Engineers can now kick off coding sessions, monitor long-running agent tasks, and interact with Kiro's spec-and-task workflow from a phone — useful for reviewing progress on autonomous tasks without being at a workstation.
The AWS DevOps Agent received a more consequential update: release management capabilities. The agent can now perform a release readiness review before a deployment, evaluating code changes against natural language standards that teams define in plain English. It can also autonomously execute change-specific tests during a release window, reducing the coordination cost of deployment sign-off. Combined with its existing code review and generation features, the DevOps Agent is beginning to cover the full inner loop from writing code through verifying it is safe to ship — a meaningful step toward what AWS is calling "release autonomy."
Additional Summit announcements relevant to developers include Amazon Bedrock enhancements with managed knowledge bases and direct web search integration for retrieval-augmented generation, AWS WAF's new AI traffic monetisation capabilities for API providers, and Amazon S3 object annotations that attach queryable metadata directly to stored objects. AWS Transform gained new features for accelerated tech debt reduction, and the Strands Agents toolkit — AWS's open-source multi-agent orchestration library — received interoperability improvements.
Read more — AWS News Blog
Amazon EC2 G7 Instances: NVIDIA Blackwell GPUs and 4.6x AI Inference Gains
AWS simultaneously launched Amazon EC2 G7 instances at the Summit, powered by NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs. The G7 generation delivers up to 4.6x AI inference performance compared to G6 instances and up to 2.1x graphics performance, making them relevant for two distinct workload categories: GPU-accelerated AI inference at scale and professional visualisation or simulation workloads.
For developers building inference pipelines or running large model serving stacks, the G7 instances represent the first Blackwell-based GPU option in EC2 without requiring the significantly higher cost and provisioning lead time of P6 instances. The RTX PRO 4500 is designed for sustained server workloads rather than consumer-grade throughput bursts, which translates to more predictable inference latency under sustained load.
AWS also noted that Graviton5 — the custom ARM processor announced at re:Invent in December 2025 — is now broadly available. With 192 ARM cores per chip and up to 25% better general compute performance, Graviton5 provides the CPU-side counterpart for teams building hybrid CPU/GPU inference architectures. The combination of Graviton5 for control-plane and orchestration logic with G7 instances for GPU inference is a reference architecture AWS is actively promoting for cost-sensitive LLM deployment.
Read more — AWS News Blog
Links & Sources
You May Also Enjoy
